analytics-ruby 2.2.3.pre → 2.2.4.pre
Sign up to get free protection for your applications and to get access to all the features.
- checksums.yaml +5 -5
- data/History.md +16 -0
- data/Makefile +17 -8
- data/README.md +2 -2
- data/RELEASING.md +2 -3
- data/Rakefile +17 -1
- data/analytics-ruby.gemspec +10 -2
- data/codecov.yml +2 -0
- data/lib/analytics-ruby.rb +1 -0
- data/lib/segment/analytics.rb +9 -2
- data/lib/segment/analytics/backoff_policy.rb +49 -0
- data/lib/segment/analytics/client.rb +148 -99
- data/lib/segment/analytics/defaults.rb +20 -4
- data/lib/segment/analytics/logging.rb +2 -4
- data/lib/segment/analytics/message.rb +26 -0
- data/lib/segment/analytics/message_batch.rb +58 -0
- data/lib/segment/analytics/request.rb +84 -32
- data/lib/segment/analytics/response.rb +0 -1
- data/lib/segment/analytics/utils.rb +19 -16
- data/lib/segment/analytics/version.rb +1 -1
- data/lib/segment/analytics/worker.rb +11 -10
- data/spec/helpers/runscope_client.rb +38 -0
- data/spec/segment/analytics/backoff_policy_spec.rb +92 -0
- data/spec/segment/analytics/client_spec.rb +61 -44
- data/spec/segment/analytics/e2e_spec.rb +48 -0
- data/spec/segment/analytics/message_batch_spec.rb +49 -0
- data/spec/segment/analytics/message_spec.rb +35 -0
- data/spec/segment/analytics/request_spec.rb +87 -34
- data/spec/segment/analytics/worker_spec.rb +24 -16
- data/spec/spec_helper.rb +32 -6
- metadata +73 -17
- data/Gemfile.lock +0 -43
- data/analytics-ruby-2.0.13.gem +0 -0
- data/analytics-ruby-2.1.0.gem +0 -0
- data/analytics-ruby-2.2.2.gem +0 -0
@@ -6,15 +6,31 @@ module Segment
|
|
6
6
|
PORT = 443
|
7
7
|
PATH = '/v1/import'
|
8
8
|
SSL = true
|
9
|
-
HEADERS = {
|
10
|
-
|
11
|
-
|
9
|
+
HEADERS = { 'Accept' => 'application/json',
|
10
|
+
'Content-Type' => 'application/json',
|
11
|
+
'User-Agent' => "analytics-ruby/#{Analytics::VERSION}" }
|
12
|
+
RETRIES = 10
|
12
13
|
end
|
13
14
|
|
14
15
|
module Queue
|
15
|
-
BATCH_SIZE = 100
|
16
16
|
MAX_SIZE = 10000
|
17
17
|
end
|
18
|
+
|
19
|
+
module Message
|
20
|
+
MAX_BYTES = 32768 # 32Kb
|
21
|
+
end
|
22
|
+
|
23
|
+
module MessageBatch
|
24
|
+
MAX_BYTES = 512_000 # 500Kb
|
25
|
+
MAX_SIZE = 100
|
26
|
+
end
|
27
|
+
|
28
|
+
module BackoffPolicy
|
29
|
+
MIN_TIMEOUT_MS = 100
|
30
|
+
MAX_TIMEOUT_MS = 10000
|
31
|
+
MULTIPLIER = 1.5
|
32
|
+
RANDOMIZATION_FACTOR = 0.5
|
33
|
+
end
|
18
34
|
end
|
19
35
|
end
|
20
36
|
end
|
@@ -0,0 +1,26 @@
|
|
1
|
+
require 'segment/analytics/defaults'
|
2
|
+
|
3
|
+
module Segment
|
4
|
+
class Analytics
|
5
|
+
# Represents a message to be sent to the API
|
6
|
+
class Message
|
7
|
+
def initialize(hash)
|
8
|
+
@hash = hash
|
9
|
+
end
|
10
|
+
|
11
|
+
def too_big?
|
12
|
+
json_size > Defaults::Message::MAX_BYTES
|
13
|
+
end
|
14
|
+
|
15
|
+
def json_size
|
16
|
+
to_json.bytesize
|
17
|
+
end
|
18
|
+
|
19
|
+
# Since the hash is expected to not be modified (set at initialization),
|
20
|
+
# the JSON version can be cached after the first computation.
|
21
|
+
def to_json(*args)
|
22
|
+
@json ||= @hash.to_json(*args)
|
23
|
+
end
|
24
|
+
end
|
25
|
+
end
|
26
|
+
end
|
@@ -0,0 +1,58 @@
|
|
1
|
+
require 'segment/analytics/logging'
|
2
|
+
|
3
|
+
module Segment
|
4
|
+
class Analytics
|
5
|
+
# A batch of `Message`s to be sent to the API
|
6
|
+
class MessageBatch
|
7
|
+
extend Forwardable
|
8
|
+
include Segment::Analytics::Logging
|
9
|
+
include Segment::Analytics::Defaults::MessageBatch
|
10
|
+
|
11
|
+
def initialize(max_message_count)
|
12
|
+
@messages = []
|
13
|
+
@max_message_count = max_message_count
|
14
|
+
@json_size = 0
|
15
|
+
end
|
16
|
+
|
17
|
+
def <<(message)
|
18
|
+
if message.too_big?
|
19
|
+
logger.error('a message exceeded the maximum allowed size')
|
20
|
+
else
|
21
|
+
@messages << message
|
22
|
+
@json_size += message.json_size + 1 # One byte for the comma
|
23
|
+
end
|
24
|
+
end
|
25
|
+
|
26
|
+
def full?
|
27
|
+
item_count_exhausted? || size_exhausted?
|
28
|
+
end
|
29
|
+
|
30
|
+
def clear
|
31
|
+
@messages.clear
|
32
|
+
@json_size = 0
|
33
|
+
end
|
34
|
+
|
35
|
+
def_delegators :@messages, :to_json
|
36
|
+
def_delegators :@messages, :empty?
|
37
|
+
def_delegators :@messages, :length
|
38
|
+
|
39
|
+
private
|
40
|
+
|
41
|
+
def item_count_exhausted?
|
42
|
+
@messages.length >= @max_message_count
|
43
|
+
end
|
44
|
+
|
45
|
+
# We consider the max size here as just enough to leave room for one more
|
46
|
+
# message of the largest size possible. This is a shortcut that allows us
|
47
|
+
# to use a native Ruby `Queue` that doesn't allow peeking. The tradeoff
|
48
|
+
# here is that we might fit in less messages than possible into a batch.
|
49
|
+
#
|
50
|
+
# The alternative is to use our own `Queue` implementation that allows
|
51
|
+
# peeking, and to consider the next message size when calculating whether
|
52
|
+
# the message can be accomodated in this batch.
|
53
|
+
def size_exhausted?
|
54
|
+
@json_size >= (MAX_BYTES - Defaults::Message::MAX_BYTES)
|
55
|
+
end
|
56
|
+
end
|
57
|
+
end
|
58
|
+
end
|
@@ -2,6 +2,7 @@ require 'segment/analytics/defaults'
|
|
2
2
|
require 'segment/analytics/utils'
|
3
3
|
require 'segment/analytics/response'
|
4
4
|
require 'segment/analytics/logging'
|
5
|
+
require 'segment/analytics/backoff_policy'
|
5
6
|
require 'net/http'
|
6
7
|
require 'net/https'
|
7
8
|
require 'json'
|
@@ -19,10 +20,11 @@ module Segment
|
|
19
20
|
options[:host] ||= HOST
|
20
21
|
options[:port] ||= PORT
|
21
22
|
options[:ssl] ||= SSL
|
22
|
-
options[:headers]
|
23
|
+
@headers = options[:headers] || HEADERS
|
23
24
|
@path = options[:path] || PATH
|
24
25
|
@retries = options[:retries] || RETRIES
|
25
|
-
@
|
26
|
+
@backoff_policy =
|
27
|
+
options[:backoff_policy] || Segment::Analytics::BackoffPolicy.new
|
26
28
|
|
27
29
|
http = Net::HTTP.new(options[:host], options[:port])
|
28
30
|
http.use_ssl = options[:ssl]
|
@@ -36,42 +38,92 @@ module Segment
|
|
36
38
|
#
|
37
39
|
# returns - Response of the status and error if it exists
|
38
40
|
def post(write_key, batch)
|
39
|
-
|
40
|
-
|
41
|
-
|
42
|
-
|
41
|
+
last_response, exception = retry_with_backoff(@retries) do
|
42
|
+
status_code, body = send_request(write_key, batch)
|
43
|
+
error = JSON.parse(body)['error']
|
44
|
+
should_retry = should_retry_request?(status_code, body)
|
45
|
+
|
46
|
+
[Response.new(status_code, error), should_retry]
|
47
|
+
end
|
48
|
+
|
49
|
+
if exception
|
50
|
+
logger.error(exception.message)
|
51
|
+
exception.backtrace.each { |line| logger.error(line) }
|
52
|
+
Response.new(-1, "Connection error: #{exception}")
|
53
|
+
else
|
54
|
+
last_response
|
55
|
+
end
|
56
|
+
end
|
57
|
+
|
58
|
+
private
|
59
|
+
|
60
|
+
def should_retry_request?(status_code, body)
|
61
|
+
if status_code >= 500
|
62
|
+
true # Server error
|
63
|
+
elsif status_code == 429
|
64
|
+
true # Rate limited
|
65
|
+
elsif status_code >= 400
|
66
|
+
logger.error(body)
|
67
|
+
false # Client error. Do not retry, but log
|
68
|
+
else
|
69
|
+
false
|
70
|
+
end
|
71
|
+
end
|
72
|
+
|
73
|
+
# Takes a block that returns [result, should_retry].
|
74
|
+
#
|
75
|
+
# Retries upto `retries_remaining` times, if `should_retry` is false or
|
76
|
+
# an exception is raised. `@backoff_policy` is used to determine the
|
77
|
+
# duration to sleep between attempts
|
78
|
+
#
|
79
|
+
# Returns [last_result, raised_exception]
|
80
|
+
def retry_with_backoff(retries_remaining, &block)
|
81
|
+
result, caught_exception = nil
|
82
|
+
should_retry = false
|
83
|
+
|
43
84
|
begin
|
44
|
-
|
45
|
-
|
46
|
-
|
47
|
-
|
48
|
-
|
49
|
-
|
50
|
-
|
51
|
-
|
52
|
-
|
53
|
-
|
54
|
-
|
55
|
-
|
56
|
-
error = body["error"]
|
57
|
-
end
|
58
|
-
rescue Exception => e
|
59
|
-
unless (remaining_retries -=1).zero?
|
60
|
-
sleep(backoff)
|
61
|
-
retry
|
62
|
-
end
|
63
|
-
|
64
|
-
logger.error e.message
|
65
|
-
e.backtrace.each { |line| logger.error line }
|
66
|
-
status = -1
|
67
|
-
error = "Connection error: #{e}"
|
85
|
+
result, should_retry = yield
|
86
|
+
return [result, nil] unless should_retry
|
87
|
+
rescue StandardError => e
|
88
|
+
should_retry = true
|
89
|
+
caught_exception = e
|
90
|
+
end
|
91
|
+
|
92
|
+
if should_retry && (retries_remaining > 1)
|
93
|
+
sleep(@backoff_policy.next_interval.to_f / 1000)
|
94
|
+
retry_with_backoff(retries_remaining - 1, &block)
|
95
|
+
else
|
96
|
+
[result, caught_exception]
|
68
97
|
end
|
98
|
+
end
|
99
|
+
|
100
|
+
# Sends a request for the batch, returns [status_code, body]
|
101
|
+
def send_request(write_key, batch)
|
102
|
+
payload = JSON.generate(
|
103
|
+
:sentAt => datetime_in_iso8601(Time.now),
|
104
|
+
:batch => batch
|
105
|
+
)
|
106
|
+
request = Net::HTTP::Post.new(@path, @headers)
|
107
|
+
request.basic_auth(write_key, nil)
|
108
|
+
|
109
|
+
if self.class.stub
|
110
|
+
logger.debug "stubbed request to #{@path}: " \
|
111
|
+
"write key = #{write_key}, batch = JSON.generate(#{batch})"
|
69
112
|
|
70
|
-
|
113
|
+
[200, '{}']
|
114
|
+
else
|
115
|
+
# If `start` is not called, Ruby adds a 'Connection: close' header to
|
116
|
+
# all requests, preventing us from reusing a connection for multiple
|
117
|
+
# HTTP requests
|
118
|
+
@http.start unless @http.started?
|
119
|
+
|
120
|
+
response = @http.request(request, payload)
|
121
|
+
[response.code.to_i, response.body]
|
122
|
+
end
|
71
123
|
end
|
72
124
|
|
73
125
|
class << self
|
74
|
-
|
126
|
+
attr_writer :stub
|
75
127
|
|
76
128
|
def stub
|
77
129
|
@stub || ENV['STUB']
|
@@ -8,7 +8,9 @@ module Segment
|
|
8
8
|
# public: Return a new hash with keys converted from strings to symbols
|
9
9
|
#
|
10
10
|
def symbolize_keys(hash)
|
11
|
-
hash.
|
11
|
+
hash.each_with_object({}) do |(k, v), memo|
|
12
|
+
memo[k.to_sym] = v
|
13
|
+
end
|
12
14
|
end
|
13
15
|
|
14
16
|
# public: Convert hash keys from strings to symbols in place
|
@@ -20,17 +22,18 @@ module Segment
|
|
20
22
|
# public: Return a new hash with keys as strings
|
21
23
|
#
|
22
24
|
def stringify_keys(hash)
|
23
|
-
hash.
|
25
|
+
hash.each_with_object({}) do |(k, v), memo|
|
26
|
+
memo[k.to_s] = v
|
27
|
+
end
|
24
28
|
end
|
25
29
|
|
26
30
|
# public: Returns a new hash with all the date values in the into iso8601
|
27
31
|
# strings
|
28
32
|
#
|
29
33
|
def isoify_dates(hash)
|
30
|
-
hash.
|
34
|
+
hash.each_with_object({}) do |(k, v), memo|
|
31
35
|
memo[k] = datetime_in_iso8601(v)
|
32
|
-
|
33
|
-
}
|
36
|
+
end
|
34
37
|
end
|
35
38
|
|
36
39
|
# public: Converts all the date values in the into iso8601 strings in place
|
@@ -42,18 +45,18 @@ module Segment
|
|
42
45
|
# public: Returns a uid string
|
43
46
|
#
|
44
47
|
def uid
|
45
|
-
arr = SecureRandom.random_bytes(16).unpack(
|
48
|
+
arr = SecureRandom.random_bytes(16).unpack('NnnnnN')
|
46
49
|
arr[2] = (arr[2] & 0x0fff) | 0x4000
|
47
50
|
arr[3] = (arr[3] & 0x3fff) | 0x8000
|
48
|
-
|
51
|
+
'%08x-%04x-%04x-%04x-%04x%08x' % arr
|
49
52
|
end
|
50
53
|
|
51
|
-
def datetime_in_iso8601
|
54
|
+
def datetime_in_iso8601(datetime)
|
52
55
|
case datetime
|
53
56
|
when Time
|
54
|
-
|
57
|
+
time_in_iso8601 datetime
|
55
58
|
when DateTime
|
56
|
-
|
59
|
+
time_in_iso8601 datetime.to_time
|
57
60
|
when Date
|
58
61
|
date_in_iso8601 datetime
|
59
62
|
else
|
@@ -61,19 +64,19 @@ module Segment
|
|
61
64
|
end
|
62
65
|
end
|
63
66
|
|
64
|
-
def time_in_iso8601
|
67
|
+
def time_in_iso8601(time, fraction_digits = 3)
|
65
68
|
fraction = if fraction_digits > 0
|
66
|
-
(
|
69
|
+
('.%06i' % time.usec)[0, fraction_digits + 1]
|
67
70
|
end
|
68
71
|
|
69
|
-
"#{time.strftime(
|
72
|
+
"#{time.strftime('%Y-%m-%dT%H:%M:%S')}#{fraction}#{formatted_offset(time, true, 'Z')}"
|
70
73
|
end
|
71
74
|
|
72
|
-
def date_in_iso8601
|
73
|
-
date.strftime(
|
75
|
+
def date_in_iso8601(date)
|
76
|
+
date.strftime('%F')
|
74
77
|
end
|
75
78
|
|
76
|
-
def formatted_offset
|
79
|
+
def formatted_offset(time, colon = true, alternate_utc_string = nil)
|
77
80
|
time.utc? && alternate_utc_string || seconds_to_utc_offset(time.utc_offset, colon)
|
78
81
|
end
|
79
82
|
|
@@ -1,7 +1,8 @@
|
|
1
1
|
require 'segment/analytics/defaults'
|
2
|
-
require 'segment/analytics/
|
3
|
-
require 'segment/analytics/
|
2
|
+
require 'segment/analytics/message'
|
3
|
+
require 'segment/analytics/message_batch'
|
4
4
|
require 'segment/analytics/request'
|
5
|
+
require 'segment/analytics/utils'
|
5
6
|
|
6
7
|
module Segment
|
7
8
|
class Analytics
|
@@ -24,10 +25,11 @@ module Segment
|
|
24
25
|
symbolize_keys! options
|
25
26
|
@queue = queue
|
26
27
|
@write_key = write_key
|
27
|
-
@
|
28
|
-
|
29
|
-
@batch =
|
28
|
+
@on_error = options[:on_error] || proc { |status, error| }
|
29
|
+
batch_size = options[:batch_size] || Defaults::MessageBatch::MAX_SIZE
|
30
|
+
@batch = MessageBatch.new(batch_size)
|
30
31
|
@lock = Mutex.new
|
32
|
+
@request = Request.new
|
31
33
|
end
|
32
34
|
|
33
35
|
# public: Continuously runs the loop to check for new events
|
@@ -37,14 +39,13 @@ module Segment
|
|
37
39
|
return if @queue.empty?
|
38
40
|
|
39
41
|
@lock.synchronize do
|
40
|
-
until @batch.
|
41
|
-
@batch << @queue.pop
|
42
|
+
until @batch.full? || @queue.empty?
|
43
|
+
@batch << Message.new(@queue.pop)
|
42
44
|
end
|
43
45
|
end
|
44
46
|
|
45
|
-
res =
|
46
|
-
|
47
|
-
@on_error.call res.status, res.error unless res.status == 200
|
47
|
+
res = @request.post(@write_key, @batch)
|
48
|
+
@on_error.call(res.status, res.error) unless res.status == 200
|
48
49
|
|
49
50
|
@lock.synchronize { @batch.clear }
|
50
51
|
end
|
@@ -0,0 +1,38 @@
|
|
1
|
+
require 'faraday'
|
2
|
+
require 'pmap'
|
3
|
+
|
4
|
+
class RunscopeClient
|
5
|
+
def initialize(api_token)
|
6
|
+
headers = { 'Authorization' => "Bearer #{api_token}" }
|
7
|
+
@conn = Faraday.new('https://api.runscope.com', headers: headers)
|
8
|
+
end
|
9
|
+
|
10
|
+
def requests(bucket_key)
|
11
|
+
with_retries(3) do
|
12
|
+
response = @conn.get("/buckets/#{bucket_key}/messages", count: 20)
|
13
|
+
|
14
|
+
raise "Runscope error. #{response.body}" unless response.status == 200
|
15
|
+
|
16
|
+
message_uuids = JSON.parse(response.body)['data'].map { |message|
|
17
|
+
message.fetch('uuid')
|
18
|
+
}
|
19
|
+
|
20
|
+
message_uuids.pmap { |uuid|
|
21
|
+
response = @conn.get("/buckets/#{bucket_key}/messages/#{uuid}")
|
22
|
+
raise "Runscope error. #{response.body}" unless response.status == 200
|
23
|
+
JSON.parse(response.body).fetch('data').fetch('request')
|
24
|
+
}
|
25
|
+
end
|
26
|
+
end
|
27
|
+
|
28
|
+
private
|
29
|
+
|
30
|
+
def with_retries(max_retries)
|
31
|
+
retries ||= 0
|
32
|
+
yield
|
33
|
+
rescue StandardError => e
|
34
|
+
retries += 1
|
35
|
+
retry if retries < max_retries
|
36
|
+
raise e
|
37
|
+
end
|
38
|
+
end
|